Further tweaks for #168

[blag]

Final tweaks for #168. This should be merged into the issue-124/exit_on_failures branch before #168 is merged in. That PR is looking pretty good, but there were a lot of places we can/should refactor, and some places we absolutely needed to refactor so we can test the new changes.

This PR:

• Adds large comment in slack_monkey_patch.js for future developers to help determine if the monkey patching is still necessary.
• Removes Object.assign() polyfill from lib/slack-messages.js - this requires node.js version > 8.3.0
• Breaks out the dummy Logger out from tests/dummy-robot.js into its own module
• Add robot_name argument to dummy Robot mock, to better mirror actual Robot constructor
• Rename api and client variables (both st2client.js instances) to api_client and auth_client in scripts/stackstorm.js
• Coalesce multiple var declarations into one
• Combine and rename hubotErrorCallback and exitProcessWithLog into LogErrorAndExit, which is now used to handle uncaught exceptions within Hubot, and is explicitly called when we cannot recover from re/connection attempts
• Refactor loadCommands to not accept an options object parameter, since we should always simply die if we can't connect or load commands
• Move promise = authenticate(); into an else block to prevent it from running while the robot is stopping
• Use array unpacking (var [var1, var2, var3] = [1, 2, 3];) since we can now that we're using Node.js > 8
• Refactor (eventsource) stream error handling to properly register a callback instead of overwriting the onerror handler
• Remove ST2 stream error handler since we now have an uncaughtException handler for hubot
• Rename eventsource's source parameter to stream, since that makes slightly more sense
• Close down the eventsource stream instead of creating a new one that immediately gets shut down
• Add a few unit tests to test if hubot is stopped if certain environment variables aren't valid